Election Machines Case Study from 2020

Is this technology worth the cost?

This is a simple & straightforward performance and cost analysis model that anyone can understand and repeat if you have access to the data. What data do you need?

• The contract listing the cost of the election machine equipment

• The system log files from all the tabulators

For this analysis, the only cost I am examining is the contract cost of the machines, support and training listed in the line items of the contract. I am ignoring all other costs like staff, precinct facility rentals etc. The actual cost of an election is higher than what I present in this study.

Here are the basic stats on the county.

In 2020, 58,362 ballots were cast.

The system logs from all 71 tabulators were examined.

The point of this chart below is that there were only a few tabulators that counted more than 600 ballots. As you can see from this graph, most of the tabulators counted less than 300 ballots.

Now let’s break this down further by system log message type. Each colored band below is a tabulator. The message types are listed on the X-Axis. The point here is that if you total up all the error messages associated with ballots NOT being counted, this is an astonishing 30%!! What happened to these not counted ballots? They were fed through the machine again until they were counted in the majority of the cases. All kinds of mischief can be the result of scanning the same ballot again and again. Use your imagination.

Here is the same data summarized differently by count and percentage.

The main question here is very simple. How do we have an election counting system that generates so main errors and requires so much “rework”? To make a simple analogy, if I were manufacturing widgets, I would have a 30% failure rate and go out of business.

Now to the cost of this system. Here is an excerpt from the contract.

The very simple analysis is that on year one, it costs $11.43 per ballot to count all ballots ($1.1MM/58,000). If we assume this equipment is good for say 6 election cycles, then the cost per ballot drops roughly to ~ $1/ballot.

Given these costs of just the equipment, can’t WE THE PEOPLE do a better job?

Given that most of the precinct tabulators count less than 400 ballots, can’t WE THE PEOPLE manage a hand count process?

Look into it in your own county and if you need some help, leave me a comment.